Re: LAST_INSERT_ID equivalent - Mailing list pgsql-general

From Erik Price
Subject Re: LAST_INSERT_ID equivalent
Date
Msg-id 3EE8D1C2.3080808@ptc.com
Whole thread Raw
In response to Re: LAST_INSERT_ID equivalent  (Ericson Smith <eric@did-it.com>)
Responses Re: LAST_INSERT_ID equivalent
Re: LAST_INSERT_ID equivalent
List pgsql-general

Ericson Smith wrote:
> While many others use currval(), we tend to grab the next ID provided by
> nextval('seq') and use that to be inserted with the record. The process
> is very atomic, and the ID is available to be used by the rest of your
> program. The only drawback is if your insert query fails there will be a
> hole in the sequence.

So you're saying that you perform a pre-query to fetch the nextval, then
you include that in your query where you perform the INSERT?  I see.
Since this is all part of the same transaction, the nextval value won't
overwrite another simultaneous INSERT, I assume.  This seems like a good
way to do it too.  I don't mind the holes in the sequence, but wouldn't
this INSERT cause the sequence to increment the primary key yet again?



Erik


pgsql-general by date:

Previous
From: Clay Luther
Date:
Subject: Choosing Between PL/PGSQL or C/C++ for Triggers/Store Procs
Next
From: Steve Crawford
Date:
Subject: Re: need a method to ping a running database